HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Models

2023/05/19 by Junyi Li, Xiaoxue Cheng, Li, Junyi +8 · 1 voice · 71 citations
Computer Science · Psychology · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning in Healthcare #Mental Health via Writing #Topic Modeling #cs.CL

paper · pdf · doi:10.48550/arxiv.2305.11747

openalex publication_date 2023/05/19 · arxiv published 2023/05/19 · openalex created_date 2023/05/24 · arxiv updated 2023/10/23 · openalex updated_date 2026/07/28

Abstract

Large language models (LLMs), such as ChatGPT, are prone to generate hallucinations, i.e., content that conflicts with the source or cannot be verified by the factual knowledge. To understand what types of content and to which extent LLMs are apt to hallucinate, we introduce the Hallucination Evaluation benchmark for Large Language Models (HaluEval), a large collection of generated and human-annotated hallucinated samples for evaluating the performance of LLMs in recognizing hallucination. To generate these samples, we propose a ChatGPT-based two-step framework, i.e., sampling-then-filtering. Besides, we also hire some human labelers to annotate the hallucinations in ChatGPT responses. The empirical results suggest that ChatGPT is likely to generate hallucinated content in specific topics by fabricating unverifiable information (i.e., about 19.5% responses). Moreover, existing LLMs face great challenges in recognizing the hallucinations in texts. However, our experiments also prove that providing external knowledge or adding reasoning steps can help LLMs recognize hallucinations. Our benchmark can be accessed at https://github.com/RUCAIBox/HaluEval.

Cited by

• TRE: Training-Free Hallucination Detection for Diffusion Language Models
• A Unified Definition of Hallucination: It's The World Model, Stupid!
• The Semantic Illusion: Certified Limits of Embedding-Based Hallucination Detection in RAG Systems
• Semantic Grounding Index: Geometric Bounds on Context Engagement in RAG Systems
• CLINIC: Evaluating Multilingual Trustworthiness in Language Models for Healthcare
• SoK: Trust-Authorization Mismatch in LLM Agent Interactions
• HalluGen: Synthesizing Realistic and Controllable Hallucinations for Evaluating Image Restoration
• Beyond Component Strength: Synergistic Integration and Adaptive Calibration in Multi-Agent RAG Systems
• Toward Honest Language Models for Deductive Reasoning
• How Much Do LLMs Hallucinate in Document Q&A Scenarios? A 172-Billion-Token Study Across Temperatures, Context Lengths, and Hardware Platforms
• ATLAS: Harnessing retrieval-augmented generation (RAG)
• MAD-Fact: A Multi-Agent Debate Framework for Long-Form Factuality Evaluation in LLMs
• Neural Diversity Regularizes Hallucinations in Language Models
• Counting Hallucinations in Diffusion Models
• Beyond Correctness: Rewarding Faithful Reasoning in Retrieval-Augmented Generation
• Abductive Preference Learning
• Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts
• Large Language Models Hallucination: A Comprehensive Survey
• Measuring Language Model Hallucinations Through Distributional Correctness
• Beyond Token Probes: Hallucination Detection via Activation Tensors with ACT-ViT
• Hallucination as an Upper Bound: A New Perspective on Text-to-Image Evaluation
• PerHalluEval: Persian Hallucination Evaluation Benchmark for Large Language Models
• Integrated Framework for LLM Evaluation with Answer Generation
• Automating Conflict-Aware ACL Configurations with Natural Language Intents
• A Novel Differential Feature Learning for Effective Hallucination Detection and Classification
• Campus AI vs. Commercial AI: How Customizations Shape Trust and Usage of LLM as-a-Service Chatbots
• EPT Benchmark: Evaluation of Persian Trustworthiness in Large Language Models
• A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models
• Prompt-Response Semantic Divergence Metrics for Faithfulness Hallucination and Misalignment Detection in Large Language Models
• Exploring Causal Effect of Social Bias on Faithfulness Hallucinations in Large Language Models
• Beyond Prompt-Induced Lies: Investigating LLM Deception on Benign Prompts
• Efficient Knowledge Probing of Large Language Models by Adapting Pre-trained Embeddings
• FAITH: A Framework for Assessing Intrinsic Tabular Hallucinations in Finance
• Hallucination to Truth: A Review of Fact-Checking and Factuality Evaluation in Large Language Models
• MultiRAG: A Knowledge-guided Framework for Mitigating Hallucination in Multi-source Retrieval Augmented Generation
• CAAD: Context-Aware Adaptive Decoding for Truthful Text Generation
• The SMeL Test: A simple benchmark for media literacy in language models
• MLP Memory: A Retriever-Pretrained Memory for Large Language Models
• Cascaded Information Disclosure for Generalized Evaluation of Problem Solving Capabilities
• BAR Conjecture: the Feasibility of Inference Budget-Constrained LLM Services with Authenticity and Reasoning
• From Queries to Criteria: Understanding How Astronomers Evaluate LLMs
• Investigating Hallucination in Conversations for Low Resource Languages
• Strategic Communication and Language Bias in Multi-Agent LLM Coordination
• Guarantees on Dynamical System Distinguishability for LLM Token Generation
• LayerCake: Token-Aware Contrastive Decoding within Large Language Model Layers
• TeamCMU at Touché: Adversarial Co-Evolution for Advertisement Integration and Detection in Conversational Search
• MALM: A Multi-Information Adapter for Large Language Models to Mitigate Hallucination
• MIRIAD: Augmenting LLMs with millions of medical query-response pairs
• Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models
• The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs
• On the Fundamental Impossibility of Hallucination Control in Large Language Models
• Reasoning Models Hallucinate More: Factuality-Aware Reinforcement Learning for Large Reasoning Models
• When GPT Spills the Tea: Comprehensive Assessment of Knowledge File Leakage in GPTs
• Unified Multimodal Understanding and Generation Models: Advances, Challenges, and Opportunities
• Robustness in Large Language Models: A Survey of Mitigation Strategies and Evaluation Metrics
• RvLLM: LLM Runtime Verification with Domain Knowledge
• Do You Keep an Eye on What I Ask? Mitigating Multimodal Hallucination via Attention-Guided Ensemble Decoding
• The Case for Repeatable, Open, and Expert-Grounded Hallucination Benchmarks in Large Language Models
• MultiHal: Multilingual Dataset for Knowledge-Graph Grounded Evaluation of LLM Hallucinations
• Campus AI vs Commercial AI: A Late-Breaking Study on How LLM As-A-Service Customizations Shape Trust and Usage Patterns
• RAG-TESTER: Automated End-to-End Testing of Retrieval-Augmented Large Language Models
• ”My AI is Lying to Me”: User-reported LLM hallucinations in AI mobile apps reviews
• GhostCite: A Large-Scale Analysis of Citation Validity in the Age of Large Language Models
• Architecting Trust in Artificial Epistemic Agents
• Meeseeks: A Feedback-Driven, Iterative Self-Correction Benchmark evaluating LLMs' Instruction Following Capability
• Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction
• ERRORQUAKE: Heavy-Tailed Error Severity Distributions in Open-Weight Large Language Models
• Toward Generalizable Evaluation in the LLM Era: A Survey Beyond Benchmarks
• HalluLens: LLM Hallucination Benchmark
• aiXamine: Simplified LLM Safety and Security
• FAIRGAME: a Framework for AI Agents Bias Recognition using Game Theory
• Sparks of Science: Hypothesis Generation Using Structured Paper Data
• C-FAITH: A Chinese Fine-Grained Benchmark for Automated Hallucination Evaluation
• HalluShift: Measuring Distribution Shifts towards Hallucination Detection in LLMs

Discussions

• HaluEval is the best one I'm aware of, but it evaluates ability to detect hallucinations arxiv.org/pdf/2305.117... [bsky, 0 points, 0 comments]

Related

• Fine-grained Hallucination Detection and Editing for Language Models
• The Dawn After the Dark: An Empirical Study on Factuality Hallucination in Large Language Models
• DiaHalu: A Dialogue-level Hallucination Evaluation Benchmark for Large Language Models
• HalluDial: A Large-Scale Benchmark for Automatic Dialogue-Level Hallucination Evaluation
• Large Language Models Hallucination: A Comprehensive Survey